Reports indicate that a 13-hour disruption in Amazon Web Services (AWS) during December stemmed from the deployment of the company's Kiro AI coding assistant, as detailed by the Financial Times based on insights from four individuals close to the situation.
The Kiro system functions as an agentic AI, capable of performing independent operations for users. Here, it apparently concluded that erasing and rebuilding the setup was necessary, which sparked the extended downtime mainly affecting operations in China.
Amazon maintains that the involvement of AI was simply coincidental and that similar problems could arise from standard developer software or human intervention. Officials attributed the incident to a mistake by the user rather than a flaw in the AI, noting that Kiro normally seeks approval prior to any steps, though the engineer in question possessed unusually extensive access rights, highlighting a problem with user permissions rather than AI independence.
Several Amazon staff members shared with the Financial Times that this marks at least the second recent instance where the firm's AI systems contributed to operational interruptions. One high-ranking AWS worker described the incidents as minor yet predictable.
Amazon introduced Kiro back in July and has encouraged its workforce to integrate the tool regularly, establishing a target of 80 percent utilization each week while monitoring uptake closely. The firm also offers the autonomous AI via a paid monthly plan to external customers.
This series of minor issues comes after a more significant AWS failure in October, which persisted for 15 hours and affected platforms including Alexa, Snapchat, Fortnite, and Venmo. Amazon cited a glitch in its automated processes as the root cause of that event.